I have two videocards which support 2 simultaneous Directdraw overlays :
Trident Blade3D (9880) AGP
SiS315 AGP
The Blade3D has singularly ugly overlays. Having 2 lame overlays instead of 1 good overlay is like, well, having 2 pennies instead of 1 nickel
This feature was also supported in older Trident chipsets (the 3DImage 9750/9850.) There were some performance problems, so the drivers disabled this feature by default.
The SiS315 has better overlays (better upscale filtering, but shrinking is just as ugly as the Blade3D.) The SiS315 was even able to use DVD-acceleration on 1 overlay (SIMC), while simultaneously playing a regular AVI/MPEG (through 2nd overlay.)
The Geforce and Radeon each support 1 hardware overlay. But both support hardware YUV->RGB BITBLT. This means either card can display motion-video (MPEG1/AVI) using the 2D BITBLT accelerator. As long as you're in a 32bpp desktop color mode, the color quality is very similar to displaying the same video with an overlay. (The overlay will still look better due to better filtering on jagged edges.) On the other hand, there is no limit to the displayable # of simultaneous BITBLT generated video-images. The upper limit is the BITBLT bandwidth and available video memory.
On the downside, from what I've seen on both the Geforce & Radeon, the BITBLT engine isn't V-SYNC aligned, so the rendered video suffers from a lot of video tearing. (Ironically, an older Rage128 and Savage3D exhibited less tearing.)